On 01/24/2014 07:08 AM, Iorga, Cristian wrote: > Hi Robert, > What is your OS? > I am unable to reproduce it on my machine following your testcase: Ubuntu > 13.10 x86 64bits. > I tried with qemux86 and qemux86-64 machines. > But then again, on my machine there is no /usr/lib/libstdc++.so file. > I assume that you have tried with that touch, see how system behaves and if > it not the case of a host contamination.
My Fedora 19 box has /usr/lib/libstdc++.so.6 and .so.6.0.19 and the recipe works for me. I guess we need to look at sed'ing the libtool script harder. The problem is the db package doesn't include enough stuff to do the usual rebuild of configure etc. Philip > /Cristian Iorga > YP > Intel > > > -----Original Message----- > From: [email protected] > [mailto:[email protected]] On Behalf Of Robert > Yang > Sent: Friday, January 24, 2014 12:19 PM > To: Paul Barker; Philip Balister > Cc: openembedded-core > Subject: Re: [OE-core] Berkeley DB fails to build > > > Seems that it is caused by the libtool, but I'm not very sure, we can > reproduce it by: > > $ sudo touch /usr/lib/libstdc++.so > $ bitbake db > > // Robert > > On 01/24/2014 03:01 AM, Paul Barker wrote: >> On 23 January 2014 17:54, Philip Balister <[email protected]> wrote: >>> On 01/23/2014 06:51 AM, Carlos Rafael Giani wrote: >>>> Hello, >>>> >>>> I noticed that since the C++ bindings for the berkeley DB were >>>> included in commit 23f29027db73d5759889b48fabdea7f6173217a7 , I >>>> cannot build db anymore. >>>> The error is: >>>> >>>> /usr/lib/libstdc++.so: file not recognized: File format not >>>> recognized >>>> >>>> this is the offending part in the log.do_compile file: >>> >>> What distro is your build machine running? I've this from people >>> using Arch Linux. >>> >> >> I've tracked this down to libtool replacing '-lstdc++' with an >> explicit reference to '/usr/lib/libstdc++.so' before it runs ld. So >> it's libtool not ld at fault here. I did look at the recipe for >> Berkeley db and spotted that it overrides do_configure, allowing the >> shipped copies of libtool and related files to be used rather than >> running the autoreconf stuff that replaces them. That's my >> understanding at least. >> >> At that point I ran away and reverted the commit in question locally >> as libtool terrifies me and I don't need the c++ bindings myself! >> >> Sorry I can't be more help but I hope that points someone in the right >> direction. >> > _______________________________________________ > Openembedded-core mailing list > [email protected] > http://lists.openembedded.org/mailman/listinfo/openembedded-core > > _______________________________________________ Openembedded-core mailing list [email protected] http://lists.openembedded.org/mailman/listinfo/openembedded-core
